Robustness & Redundancy with regard to computer kit is sometimes challenging!
Recently I had a good slap in the face with a wet fish 🙁 When it comes to backing up my data, I’m a paranoid Pete and of course my kit is protected by UPS in case of power cuts. I make regular disk images of the boot drive and settings etc so I’m well prepared for “disasters” or so I thought 🙁
My Fujitsu Econel 200 S2 server (purchased in 2006) has given me trouble free service for years. Originally it had Windows 2003 server OS and for the past few years it has been running CentOS (now 7). Unfortunately, the PSU died and I just could not source a new or even a used one. That was something I hadn’t anticipated, the best option was to get another server. New ones are not in my budget but a search on eBay found a Fujitsu Primergy TX140 S1p Tower Server, with 8GB RAM, a redundant power supply and it came with 3 hot swap caddies. It was in my budget so I wasted no time in ordering it.
On arrival, my world was turned upside down. I hadn’t come across (server) hardware this new for a few years and was literally stumped! I was seriously out of date 🙁 The server had a redundant power supply (it has 2 but uses only one), I was familiar with that but it had 4 x 3.5 drive bays and 3 hot swap caddies. I have come across hot swap caddies but wasn’t really up to speed on the “workings”. The Mobo had a PCI raid card which fed the backplane for the 4 drive bays and only had 2 x SATA connections, to be used for “optical” drives.Seriously now out of my comfort zone.
Normally I would mirror the “boot /system” drive and then just backup the 2 x 1TB data drives to a USB HDD, this clearly wasn’t going to be possible. The TX140 only accepts 3.5 HDDs and these have to be attached to the RAID card (via the backplane) using the hot swap caddies. The RAID is configured by a a very strange “WinBios”, I tried to just use RAID 0 so I could have 3 separate drives but it didn’t like it.
I decided to forgo the mirrored drive and hung the boot drive on the “optical” drive cable. I then had 3 x hot swap caddies so decided to purchase 3 x new I TB drives and create a raid 5 array for the data, much better than I had before.
Note! The TX140 is available with a backplane that can use upto 8 x 2.5 HDDs!
So from a robustness and redundancy point of view I now have a pretty good set up. If the the PSU fails, I still have a “spare” but plan to purchase a “spare spare”. From a data point of view, I have a RAID 5 array with hot swap-able drives so if one drive fails, I can just replace the drive and rebuild the array with no loss of data and I backup the data array (using rsync) to a USB drive. In addition to this I also purchased an extra USB drive for a spare Windows box and use Robocopy to “mirror” the USB drive on CentOS to the Windows USB drive.
I’m still working on the “mirror” option for the CentOS system drive but for now I’ll just make regular live disk images using MondoRescue. I think the best option would be to install another PCI RAID card but for now, I’m reasonably happy that my data is looked after.
Robustness and redundancy are things that need to be re-assessed on a regular basis and you have to accept that things change. Complacency is your (my) downfall.
If you’re running a home or business server CentOS and Linux in general are fantastic open source options and are free but they are evolving and just getting better. You need to keep up-to-date, things change! It seems I know less about Linux now than I ever did 🙁 Check out systemd and the new way of using network tools, you no longer need to install “network Tools” “ifconfig” is old hat ip-add is the new way. Fdisk & Parted are now old hat, replaced with Gdisk. It’s been back to school for a few weeks and CentOS is very forgiving, it copes well with idiots looking after it 🙂